Audio Recorder — Windows audio recorder

The original need was straightforward: keep the audio of a prospect call without the weight, complexity and extra personal data of a video capture. Meeting platforms and screen recorders can often save an entire session, but they are oversized when a compact speech recording is all that is required.

Audio Recorder does not capture the screen. The user chooses the application playing the call, decides whether to include the microphone and starts recording. The output is a compact MP3 that works with standard players and can later be archived or passed to a transcription tool. The interface also reminds users to inform the other participant and obtain consent before recording.

At launch, the interface lists audio-enabled Windows applications such as Google Chrome, Discord or a meeting client. Search makes it easier to identify the right process. Selecting one source keeps the intended conversation separate from unrelated computer sounds.

The microphone option adds the user's voice to the same file. When recording stops, the completed MP3 is stored in Documents\AudioRecorder, and a dedicated button opens that folder. At mono 64 kbps, the interface estimates roughly 29 MB per hour, a practical compromise for speech and far smaller than a video recording.

The user interface uses React and TypeScript, while Tauri 2 connects the desktop window to the Rust engine. The native layer uses Windows audio capabilities to capture the selected process and, when requested, mix it with microphone input. Processing stays on the computer and the application requires neither an account nor a remote service.

This architecture produces a real Windows installer without bundling a complete browser. Tauri permissions were narrowed to the required commands, the content security policy was hardened and dependencies were updated. Recordings are first written to a temporary .part file and renamed only after successful finalization, reducing the chance that an interrupted file appears complete.

The starting point is apecia-org/audio-recorder, an MIT-licensed GitHub project. I forked it to my personal account, audited it, adapted it and maintained it for this specific workflow. The upstream repository and licence remain credited in the source and downloadable release.

The improvements include a call-focused French interface, explicit local-storage information, a direct output-folder action, file-size guidance, reduced permissions, hardened application settings and more reliable saving. The work demonstrates a valuable engineering capability: assessing an existing open-source foundation, checking its trust boundaries and turning it into a coherent tool for a real operational need.

The reviewed version contains no recording upload, telemetry, online account or remote updater. Audio stays in the displayed local folder. Automated checks cover the JavaScript dependency audit, frontend build, Rust formatting, tests and Clippy. Real MP3 outputs were also opened and inspected after a local recording test.

The application is currently distributed without a code-signing certificate, so Windows SmartScreen may warn on first download. That warning is not proof of malware, but the missing signature must be stated clearly. The public source lets users inspect and build the same project themselves. Users remain responsible for obtaining the consent required to record a conversation and protecting the resulting files.

Version 0.1.1 is available from my personal GitHub account as an EXE installer and an x64 Windows MSI package. The release page keeps the binaries, release notes, upstream attribution and known limitations together, making it the recommended download location instead of an opaque standalone file.

The fork remains public for code review and future changes. This is not a cloud service: it runs locally on Windows and targets calls where a compact audio record is enough. Before an important conversation, a short test of the selected application, microphone and completed MP3 is still recommended.